PDA

View Full Version : 24 Bit ADC Output Flickering



hcwong
06-14-2005, 06:29 AM
I am currently working on a project to measure with an accuracy of 1 micro volt signal within the range of·- 2.5V to·+ 2.5V with the data rate of 60 samples/sec.·From my calculation,·the effective resolution I require for this application is at least 22 bits if my·ADC·input range is - 2.5V to + 2.5V. I had tried using·a 24-bit·ADC AD7712AN from Analog Devices for this application. The problem is I only get a effective resolution of 18 bits because the the lower 6 bits is flickering. Even I ground both the positive and negative input of the AD7712AN, the lower 6 bits is not zero and values keep flickering.

Is there anyway to remove this flickering problem? Is there any other low noise ADC available in the market which does not have this problem?···

Paul Baker
06-15-2005, 02:05 AM
hcwong said...
I am currently working on a project to measure with an accuracy of 1 micro volt signal within the range of·- 2.5V to·+ 2.5V with the data rate of 60 samples/sec.·From my calculation,·the effective resolution I require for this application is at least 22 bits if my·ADC·input range is - 2.5V to + 2.5V. I had tried using·a 24-bit·ADC AD7712AN from Analog Devices for this application. The problem is I only get a effective resolution of 18 bits because the the lower 6 bits is flickering. Even I ground both the positive and negative input of the AD7712AN, the lower 6 bits is not zero and values keep flickering.

Is there anyway to remove this flickering problem? Is there any other low noise ADC available in the market which does not have this problem?
Your problem isn't with the ADC but how you are using it. High Precision (>16bit) ADCs require special consideration for thier proper operation. You need to be well versed in the specification of the ADC to get it to work properly (it is availible here (http://rocky.digikey.com/WebLib/Analog%20Devices/Web%20Data/AD7712.pdf)). You are likely experiencing digital noise caused by your stamps internal transistors switching off and on. The analog supply and digital supply must appear as electrically isolated to the ADC (this is done by having bypass capacitors near the chip's analog and digital supply), look at page 19 for a discussion of the topic. Be sure the ground of your ADC and the stamp are tied together. And finally try to match the input impedance to the ADC (page 15) since reflections due to mismatched impedances will skew an oscillating input. In a nutshell, read the spec back and forth until you understand, if after reading the spec and searching google for explainations of concepts you still don't understand something, ask on the forum and either I or someone else will do our best to explain it.

Jonathan
06-15-2005, 02:47 AM
Are you working on a breadboard? It is very difficult to get stable readings on a breadbaord with precision ADCs. However, everything in the post above is VERY true, without a really stable Vref and plenty of filtering you will get flakey readings regardless of mounting.

Jonathan

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.madlabs.info (http://www.madlabs.info) - Home of the Hydrogen Fuel Cell Robot

micropar
06-15-2005, 03:04 AM
Hi Wong,

What if, if you increase your measuring range from -12v to +12v and then use OP-AMP to scale it down to AD convertor range, keeping OP-AMP very near to BS?

Regards,
--micropar--

hcwong
06-16-2005, 08:53 PM
Dear Paul,

Sorry for not replying as I was travelling.

Many thanks for your advise. I had already·read the pages·in detail and tested varous methods but the problem·cannot be eliminated.

I am currently building the project on a vero board and not using a breadboard.

I am using 2 different power supplies, i.e. for analog power and digital power. The ground of both the analog power and digital power is linked with one copper trace as advised in the specs. I had used sufficient coupling capacitors (10uF and 0.1uF) on the power supplies pin of ADC as advised in the specs.

I had also used a separate Vref IC and not using the in-build Vref as the in-build Vref is known to have noise.

I had connected the ground of ADC and BS2 together to ensure both circuits have a common ground.

I had also used anti aliasing filters but the problem still persist. I may not use the correct capcitor and resistor for the anti aliasing filter. Any idea how to calculate and use the correct capacitor and resistor for the anti aliasing filter?

Any other suggestions please?


Paul Baker said...

hcwong said...
I am currently working on a project to measure with an accuracy of 1 micro volt signal within the range of·- 2.5V to·+ 2.5V with the data rate of 60 samples/sec.·From my calculation,·the effective resolution I require for this application is at least 22 bits if my·ADC·input range is - 2.5V to + 2.5V. I had tried using·a 24-bit·ADC AD7712AN from Analog Devices for this application. The problem is I only get a effective resolution of 18 bits because the the lower 6 bits is flickering. Even I ground both the positive and negative input of the AD7712AN, the lower 6 bits is not zero and values keep flickering.

Is there anyway to remove this flickering problem? Is there any other low noise ADC available in the market which does not have this problem?
Your problem isn't with the ADC but how you are using it. High Precision (>16bit) ADCs require special consideration for thier proper operation. You need to be well versed in the specification of the ADC to get it to work properly (it is availible here (http://rocky.digikey.com/WebLib/Analog%20Devices/Web%20Data/AD7712.pdf)). You are likely experiencing digital noise caused by your stamps internal transistors switching off and on. The analog supply and digital supply must appear as electrically isolated to the ADC (this is done by having bypass capacitors near the chip's analog and digital supply), look at page 19 for a discussion of the topic. Be sure the ground of your ADC and the stamp are tied together. And finally try to match the input impedance to the ADC (page 15) since reflections due to mismatched impedances will skew an oscillating input. In a nutshell, read the spec back and forth until you understand, if after reading the spec and searching google for explainations of concepts you still don't understand something, ask on the forum and either I or someone else will do our best to explain it.

hcwong
06-16-2005, 08:56 PM
I had tried this as well. I am currently amplifying the signal 6x and using the maximum available full-scale input of the ADC.

micropar said...
Hi Wong,

What if, if you increase your measuring range from -12v to +12v and then use OP-AMP to scale it down to AD convertor range, keeping OP-AMP very near to BS?

Regards,
--micropar--

Paul Baker
06-17-2005, 12:21 AM
Ah so your wirewrapping, that may be the issue. The long pins of the sockets introduce parasitic capacitance but more importantly parasitic inductance. But the biggest problem with getting the full precision is you are not providing an adequate analog ground, this is absolutely critical for the proper operation of the ADC. Read this article (http://www.sigcon.com/Pubs/edn/adcgrounding.htm)on the issue of analog grounds in a digital environment. To ensure proper operation you will likely need to put the chip on a PCB.

hcwong
06-22-2005, 07:26 AM
The problem is I am not using wirewrapping. I am actually soldering all my components onto the Vero board.

Any more ideas how to improve?


Paul Baker said...
Ah so your wirewrapping, that may be the issue. The long pins of the sockets introduce parasitic capacitance but more importantly parasitic inductance. But the biggest problem with getting the full precision is you are not providing an adequate analog ground, this is absolutely critical for the proper operation of the ADC. Read this article (http://www.sigcon.com/Pubs/edn/adcgrounding.htm)on the issue of analog grounds in a digital environment. To ensure proper operation you will likely need to put the chip on a PCB.

Jon Williams
06-22-2005, 07:55 AM
I had lunch with a salesman for a large electronics company last week and in jest about 24-bit ADCs he said, "Perhaps 20 bits are usable [under pristine conditions] ... the rest are for marketing." The point being that unless you've got a top-notch board design and layout you're probably not going to get rid of the noise in the lower bits.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Jon Williams
Applications Engineer, Parallax

Tracy Allen
06-22-2005, 11:23 PM
Imagine a stack of 2^24=16777216 sheets of paper. A ream of paper contains 500 sheets and is about 2" thick. Your stack will be 33554 reams high, or about 67109 inches, or 5592 feet, or a little over one mile. You want to measure to the accuracy of one sheet of paper in that stack!

The optimal use of these converters involves a circuit board -
-- with the sensor element at near zero distance from the ADC inputs, that is ADC integrated with the sensor
-- differential inputs
-- inputs with a guard ring
-- multilayer printed circuit board with power plane layers
-- all the digital traces run far from the analog, and against the power plane
-- heavy power filtering
-- probably inside a shielding enclosure

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Tracy Allen
www.emesystems.com (http://www.emesystems.com)

Jon Williams
06-22-2005, 11:31 PM
Great analogy!

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Jon Williams
Applications Engineer, Parallax

Chris Savage
06-22-2005, 11:34 PM
Jon,

·· That's what I was thinking!· After I sat there in bewilderment for a few moments!· Way to go Tracy!

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chris Savage
Parallax Tech Support
csavage@parallax.com (mailto:csavage@parallax.com)

Paul Baker
06-23-2005, 12:03 AM
You soldering components onto the vero board does not address the analog/digital ground issue. Also if you are not using a true differential interface you will loose precision. What Tracey and Jon said is very true, to come close to that degree of precision (if at all possible) takes a special expertise in board layout which only comes through trail and error gained from many years of experience. A pure digital designer can ignore some aspects nessesary to analog design, and an analog designer can ignore some aspects required for digital design. But ADCs are both analog and digital in the same chip and no aspect of design can be ignored if you intend on getting the most out of your design.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

hcwong
07-02-2005, 06:22 AM
Thanks for your comments.

Actually my input is a differential input.

As my knowledge in analog electronics design is limited, I do not know how to go about designing the circuit board with the following features as mentioned earlier by Tracy.

-- inputs with a guard ring
-- multilayer printed circuit board with power plane layers
-- all the digital traces run far from the analog, and against the power plane
-- heavy power filtering
-- probably inside a shielding enclosure

Is there any book or example circuit board drawing that I can use as a guideline?·Are there any expert in this forum that can help me design the circuit board?

Another thing, how do I provide heavy filtering·to the·power supply? I an currently using a transformer, a·voltage regulator IC, a few diodes and a few capacitors for·both analog and digital power supply. I have two of the same power supply, one for digital power and another for analog power.

Thanks in advance.

Mohamed Refky
07-04-2005, 12:20 AM
Analog Devices offer evaluation boards to assist cusotomers in theire initial layout.
The layout of evaluation board is optimized in terms of grounding,decoupling and
signal routing and can be used as a model when laying out your board.

Published by Analog Devices:
Practical Design Techniques For Sensors Signal Conditioning

Look at the AD7712 Technical Note :
http://www.analog.com
AD7712 / Evaluation Boards

page 7 for silkscreen and page 8 for the layout picture of the evalutation board.

I use AD7730 for weigh scale application,I use single supply for AVdd and DVdd
( 7805 ) voltage regulator.

You should rout the layout on double sided board with ground plane as the EVAL board.

Here is some comments from Analog Devices support:

* Put ground plane under the AD77XX package.
* Connect AGND pin and DGND pin to this plane.
* Connect analog signals path and power supply path separtely to this ground.
* Locate analog circuits and digital circuits separated as possible.
* Metal box could be necessary -depends on environment and should be connected
to ground plane.

Regards
Mohamed Refky

Tracy Allen
07-04-2005, 04:20 AM
The application notes that Mohamed suggested would be an excellent way to start. Also, application notes from other manufactures, Linear Tech (LTC24xx series) TI/Burr-Brown (ADSxxx series). if you don't have it already, The Art of Electronics (Horowitz & Hill) offers a lot of insight into low noise and precision techniques from guys who build astronomy instrumentation.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Tracy Allen
www.emesystems.com (http://www.emesystems.com)